Add ReadOnlySequence<T>.FirstSpan

[maryamariyan]

Fixes: #33029

• Expose API
• Add unit tests
• Add perf tests

cc: @ahsonkhan

[ahsonkhan]

@maryamariyan, consider making this a draft PR while in progress. This way, the WIP/No merge label become unnecessary. Something to consider for the future.

See https://github.blog/2019-02-14-introducing-draft-pull-requests/

[maryamariyan]

I added benchmarks for FirstSpan in the new performance repo. (dotnet/performance#410 (comment))
Below shows comparison between using First.Span vs the new FirstSpan API

|                  Method |     Mean |     Error |    StdDev |   Median |      Min |      Max | 
|------------------------ |---------:|----------:|----------:|---------:|---------:|---------:|
| First.Span for Array    | 28.58 ns | 0.3199 ns | 0.2835 ns | 28.52 ns | 28.22 ns | 29.33 ns |
| FirstSpan for Array     | 20.11 ns | 0.7130 ns | 0.8211 ns | 19.68 ns | 19.29 ns | 22.48 ns |

| First.Span for Memory   | 29.21 ns | 0.1973 ns | 0.1648 ns | 29.20 ns | 28.93 ns | 29.49 ns |
| FirstSpan for Memory    | 21.75 ns | 0.4870 ns | 0.4317 ns | 21.62 ns | 21.30 ns | 22.86 ns |

| First.Span for 1 Segment  | 26.07 ns | 0.5792 ns | 0.5418 ns | 25.92 ns | 25.59 ns | 27.39 ns |
| FirstSpan for 1 Segment   | 25.38 ns | 0.8425 ns | 0.9015 ns | 24.97 ns | 24.54 ns | 27.50 ns |

| First.Span for 10 Segments | 26.05 ns | 0.9496 ns | 1.0161 ns | 25.73 ns | 25.13 ns | 28.96 ns |
| FirstSpan for 10 Segments  | 23.69 ns | 0.1279 ns | 0.0999 ns | 23.67 ns | 23.57 ns | 23.86 ns |

|First.Span for MemoryManager| 40.65 ns | 0.9715 ns | 0.9542 ns | 40.39 ns | 39.63 ns | 42.59 ns |
| FirstSpan for MemoryManager| 28.41 ns | 0.7400 ns | 0.6922 ns | 28.16 ns | 27.85 ns | 30.22 ns |

| First.Span for String | 26.21 ns | 0.1975 ns | 0.1542 ns | 26.20 ns | 25.90 ns | 26.44 ns |
| FirstSpan for String  | 23.59 ns | 1.1345 ns | 1.3065 ns | 22.98 ns | 22.52 ns | 26.28 ns |

[ahsonkhan]

Otherwise, LGTM

[ahsonkhan]

I added benchmarks for FirstSpan in the new performance repo

How about for string and MemoryManager cases?

[maryamariyan]

I just updated my comment to also include numbers for string and MemoryManager

[maryamariyan]

Linux failure related to Fedora 27 queues leaving Helix.